Chemical Geology xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Chemical Geology
journal homepage: www.elsevier.com/locate/chemgeo

Lu-Hf ratios of crustal rocks and their bearing on zircon Hf isotope model
ages: The effects of accessories
⁎
F. Bea , P. Montero, J.F. Molina, J.H. Scarrow, A. Cambeses, J.A. Moreno
Department of Mineralogy and Petrology, University of Granada, Campus Fuentenueva, 18002 Granada, Spain

A R T I C L E I N F O A B S T R A C T

Keywords: All other factors being equal, the calculation of zircon Hf two stage model ages (TDM Hf) depends on the par-
Zircon geochronology ticular Lu/Hf value assumed for the magmatic source, the effect being more pronounced as the age difference
Zircon Hf model ages between zircon and magmatic source increases. It is generally considered that the Lu/Hf measured in the zircon-
Lu/Hf crustal source hosting rock does not represent the composition of the source because of potential garnet or zircon fractionation.
Crust formation
Accordingly, most authors either assume a single fixed value for Lu/Hfsource, often Lu/Hf ≈ 0.079 to 0.108, or
use two alternative models, one for felsic sources, often Lu/Hf ≈ 0.09, and the other for mafic sources, often Lu/
Hf ≈ 0.165. In contrast with these opinions, however, here we show that partial melting of peraluminous
sources causes little decoupling of Lu from Hf because of similar solubilities of zircon and monazite.
Furthermore, the effects of residual garnet are largely compensated by the numerous zircon inclusions that
garnet and other residual minerals almost always contain. Partial melting of metaluminous sources may sig-
nificantly decouple Lu from Hf if allanite and/or titanite are not present in the source, but the effect decreases as
the melt fraction increases. Similarly, fractional crystallization of metaluminous magmas may decouple Lu from
Hf if amphibole or clinopyroxene begin to crystallize before zircon saturation. The Lu/Hf distribution in 4784
rocks from different regions and ages is lognomal rather than normal, and the calculated medians, i.e. the
maximum of the probability density function for the logarithmically transformed Lu/Hf, are Lu/Hfmafic
rocks ≈ 0.08, Lu/Hffelsic rocks ≈ 0.05, i.e. notably lower than the above-mentioned felsic and mafic magmatic
source averages. Magmatic sources may be remarkably heterogeneous with respect to Lu/Hf. Our calculations
show that fixed Lu/Hfsource values translate the Lu/Hf heterogeneity of the source to the TDM Hf thus producing
an artificial distribution of model ages that may be erroneously interpreted as different episodes of crustal
growth. Therefore, we propose that the best strategy to calculate two stage Hf model ages of zircon is to use the
analytically determined whole-rock Lu/Hf ratio as a proxy of the source. In the case of detrital or inherited
zircons, for which no whole-rock information is available, it is advisable first to determine whether they come
from a mafic or felsic rock by interpreting cathodoluminescence images, Th/U ratios and other chemical
parameters, and then venture an estimate of the Lu/Hfsource from the SiO2 average.

1. Introduction
As a result of the worldwide adoption of the LA-MC-ICPMS tech-
nique (e.g., Thirlwall and Walder, 1995; Griffin et al., 2002; Woodhead
et al., 2004) the analysis of Hf isotopes in zircon has become increas-
ingly popular for studying the formation of the crust and the petro-
genesis of granite rocks. Unfortunately, many researchers often forget,
or are unaware, that the proper interpretation of zircon Hf isotope data
has several stringent requirements which, if not followed, may result in
wrong conclusions (see Payne et al., 2016, and Vervoort and Kemp,
2016, for discussion). The calculation of two stage Hf model ages (TDM
Hf) has four main sources of uncertainty: (1) choice of the decay
constant (2) choice of mantle curve (3) estimation of the U-Pb zircon
crystallization age (4) choice of the Lu/Hf of the crustal source. Several
authors have debated points 1 to 3 quite thoroughly (e.g., Vervoort and
Patchett, 1996; Blichert-Toft and Albarede, 1997; Blichert-Toft and
Albarède, 2008; Patchett et al., 2004; Dhuime et al., 2011; Kemp et al.,
2007; Dhuime et al., 2011; Vervoort and Kemp, 2016; Iizuka et al.,
2017) but the Lu/Hf composition of potential crustal sources has rarely
been discussed, particularly from the perspective of Lu and Hf residence
in accessory minerals.
Two stage Hf model ages require assumption of a given crustal
source Lu/Hf composition to trace the 176Hf/177Hf evolution from the
zircon U-Pb crystallization age to interception with the selected mantle

⁎
Corresponding author.
E-mail address: fernando_bea@mac.com (F. Bea).

https://doi.org/10.1016/j.chemgeo.2017.11.034
Received 15 June 2017; Received in revised form 17 November 2017; Accepted 23 November 2017
0009-2541/ © 2017 Elsevier B.V. All rights reserved.

Please cite this article as: Bea, F., Chemical Geology (2017), https://doi.org/10.1016/j.chemgeo.2017.11.034
F. Bea et al. Chemical Geology xxx (xxxx) xxx–xxx

4000
Zr
LREE
P

Fig. 1. Zircon Hf model ages are calculated in two steps: first, tracing the 176Hf/177Hf
evolution from time = 0 to the zircon U/Pb crystallization age; second, tracing the
176
Hf/177Hf evolution from this point until interception with the selected mantle curve.
Note the dependence of TDM Hf from the Lu/Hf assumed for the magmatic source.
A) Archean source
4500
500 1000
ppm dissolved
100
50
700 800 900 1000
T (oC)
Fig. 3. Solubility curves of zircon (Boehnke et al., 2013), monazite (Stepanov et al., 2012)
and apatite (Harrison and Watson, 1984, modified by Bea et al., 1992 for peraluminous
melts) in felsic systems as functions of temperature represented by the concentration in
the melt of Zr, LREE and P, respectively. The curves for monazite and zircon are nearly
parallel and adjacent one to the other. The curve for apatite is steeper. Considering the
concentrations of Lu and Hf in the three accessories, these curves explain the lack of
significant Lu/Hf fractionation in peraluminous melts. See text.
curve (Fig. 1). This assumption is undoubtedly the weakest point of
model age calculations because of the lack of a robust basis for choosing
a particular Lu/Hf value and the strong influence of this on the resulting
zircon Hf model ages (TDM Hf), particularly if zircons are much younger

.y.

than the magmatic sources (Fig. 2). It is generally considered that the
20 0 m

00 .y.
.y.
m

bulk composition of the zircon-hosting rock is not a suitable proxy for

0

m
00
25

.y.
Hf model age (TDM Hf)

the Lu/Hf of the source because intracrustal differentiation processes

=

m
15

0
=
4000

00 involving garnet or zircon might significantly decouple the two ele-

=

1 .y.
= ments (but see below). This consideration and the lack of evidence
0 0m
=5 about the primary host-rock composition of inherited or detrital zircons
have led many authors to assume a fixed Lu/Hf value supposed to re-
3500

=0 present the magmatic source. Such values include the average value for
potential crustal protoliths (e.g., Amelin et al., 1999), the average upper
continental crust (Bodet and Schärer, 2000; Iizuka et al., 2005) or the
F.S. M.S. average bulk continental crust (Griffin et al., 2002; Hawkesworth and
3000

Kemp, 2006. However, given that Lu/Hf decreases significantly from

0.03 0.06 0.09 0.12 0.15 0.18 mafic to felsic rocks (see Section 5), most authors now assume two
Lu/Hf (element ratio) values, one for felsic to intermediate magmatic sources: generally Lu/
Hf ≈ 0.079 to 0.108 depending on the crustal model selected, and the
B) Paleoproterozoic source other for mafic magmatic sources: generally Lu/Hf ≈ 0.13 to 0.21, Lu/
4500

Hf 0.165 is the most widely employed.

Nevertheless, these assumptions cannot be taken for granted and
4000
Hf model age (TDM Hf)

need more consideration. Firstly, it is not clear whether intracrustal

.y.
m . granite magmatism may lead to large Lu/Hf variations. As discussed in
00 .y
3500

15 m Section 4 the similar behavior of Lu-rich and Hf-rich accessories during

= 200
1 y. partial melting and crystallization tends to mutually cancel their effects
= m.
00 on Lu/Hf. Secondly, the above-mentioned “magmatic source averages”
3000

9 .
= m.y
00 are different from many published high-quality Lu/Hf data of crustal
=6
.y.
0m rocks from which abundant granite magmas have been formed. For
= 30
2500

example, Vervoort and Patchett (1996) studied Precambrian granites

=0 M.S. and found a distribution skewed to higher values with a mode of Lu/
F.S.
2000

Hf = 0.025 and arithmetic mean of Lu/Hf = 0.066; Moyen and Martin

0.03 0.06 0.09 0.12 0.15 0.18 (2012) compiled a database of Archean TTGs with an arithmetic mean
Lu/Hf (element ratio) of Lu/Hf = 0.035; Bea et al. (2003) found that the magmatic sources of
Central Iberian granites cluster around Lu/Hf = 0.04, etc.
Fig. 2. Simulations of how the Lu/Hf selected for the source affects the TDM Hf of zircons These considerations motivated us to study the variations of Lu/Hf
with different U/Pb age from Archean (A) and Paleoproterozoic (B) sources. F.S. and M.S. in common crustal rocks and their influence on zircon Hf model ages,
represent the commonly used felsic source and mafic source averages, respectively. Δ
summarizing here the most important results. The strategy adopted was
represents the age difference between the zircon and its source. Note how the influence of
Lu/Hf is greater as it and Δ increase. as follows: Firstly, we used available textural evidence and experi-
mental data on the solubility of accessory phases in crustal melts to
model the behavior of Lu and Hf during partial melting and melt seg-
regation. Secondly, we compiled a large set of whole-rock analysis of
Archean to Mesozoic igneous rocks from Eurasia, Africa, America, and

2
F. Bea et al.
Australia. Using the dataset we studied whether Lu/Hf varies system-
atically with rock type and silica contents, and which Lu/Hf values are
likely to yield the most accurate TDM Hf for a given rock type. Lastly, we
checked the results against several well-studied natural examples and
discussed the best approach to calculate zircon Hf model ages.
2. Samples and methods
The mineral dataset used here comprises the LA-ICP-MS analyses
described in Bea (1996a) and Bea et al. (2006b). The whole-rock da-
taset comprises 4784 rocks analyzed with NAA or ICP-MS. About 80%
of them were performed by the authors at the University of Granada on
samples from Iberia, the Urals, Kola, Transbaikalia, Ivrea-Verbano, the
Arabian-Nubian Shield, the Neoproterozoic to Archean inliers of the
Egypt Western Desert, the Sanandaj-Sirjan zone in Iran, and the Ar-
chean to Proterozoic and Early Paleozoic terranes from the Reguibat
and the Reguibat Promontory. The remaining 20% were collected from
the literature and include the above-mentioned TTG database of Moyen
and Martin (2012), as well as samples from the Cordilleran batholiths of
America, the Leo Rise of the Western African Craton, the Hoggar,
Amazonia, China, Tibet, and the Lachland Fold Belt. The datasets used
are available from the authors upon request.
It is important to note that ICP-MS analysis may underestimate Hf
concentrations depending on the sample solution method used (Hall
and Plant, 1992; Totland et al., 1992; Park et al., 2013), hence pro-
ducing erroneously high Lu/Hf ratios. Hydrofluoric evaporation at at-
mospheric pressure does not fully dissolves the zircon grains present in
a sample. Hydrofluoric attack at elevated pressure either in Parr-type
3
Chemical Geology xxx (xxxx) xxx–xxx
Fig. 4. Evolution of Lu and Hf and Lu/Hf as functions of the
melt fraction in peraluminous systems (see Appendix for de-
tails). Despite the marked variations in the concentrations of
the two elements, Lu/Hfmelt is nearly constant and close to Lu/
Hsource. because of the similar effect of temperature on the
solubility of monazite (Lu main carrier) and zircon (Hf main
carrier) as shown in Fig. 3.
bombs or high-pressure microwave reactors often dissolves all zircon;
nonetheless, the ICP-MS Zr results must be compared with Zr XRF re-
sults before they may be accepted. Alkaline fusion prior to acid attack
dissolves all zircon. To avoid outliers, therefore, the initial dataset was
filtered to exclude: (1) samples analyzed with ICP-MS after low-pres-
sure dissolution, (2) samples with Lu/Hf > 0.5 because these are likely
caused by incomplete zircon dissolution or reflect ratio errors caused by
extremely low Hf concentrations, (3) samples with Zr/Hf > 60 or Zr/
Hf < 10 because they probably reflect anomalous rocks or wrong
analyses (see Bea et al., 2006b), (4) samples with SiO2 < 45 wt% or
SiO2 > 76 wt% to avoid oddities, metasomatic and extremely frac-
tionated rocks.
To calculate the Hf isotope parameters, we used the following CHUR
values 176Lu/177Hf = 0.0336; 176Hf/177Hf = 0.282785 (Bouvier et al.,
2008), and λ 176Lu = 1.867 ∗ 10− 11 y− 1 (Scherer et al., 2001;
Söderlund et al., 2004). We also assumed the mantle source was de-
pleted and changed from chondritic (εHf = 0) at 4.56 Ga to εHf0Ga =
+ 16 today (but see Vervoort and Kemp, 2016; and references therein
for discussion). To calculate the Nd isotope parameters we used the
presently accepted CHUR Sm–Nd reference parameters,
147
Sm/144Nd = 0.1966 and 143Nd/144Nd = 0.512638 (Jacobsen and
Wasserburg, 1980, 1984), and assumed a depleted mantle that changed
from chondritic (εNd = 0) at 4.56 Ga to εNd = + 10 at the present day
(Goldstein et al., 1984).
3. Crustal residence of Lu and Hf
The crustal geochemistry of Lu and Hf is mainly controlled by
F. Bea et al.
Fig. 5. Textural aspect of granulites: (A) mafic granulite from Ivrea-Verbano, Italy (B)
felsic granulite from the Lapland Granulite Belt, Russia. Note the abundance of zircon
included in garnet, pyroxene, and plagioclase. If melts were extracted from these rocks,
the effects of residual garnet on Lu/Hf ratios would be largely compensated by the re-
tention of zircon; see Table 1. Mineral abbreviations are from Kretz (1983).
Table 1
Average concentrations and Lu/Hf in seven leucosome/melanosome pairs from the Ivrea-
Verbano metapelitic granulites. Despite the melanosomes contain > 40 vol% modal
garnet, the Lu/Hf in melanosomes and leucosomes is virtually identical. This happens
because residual garnet contain numerous zircon inclusions that compensate the effects of
garnet.
Lu (ppm) Hf (ppm) Lu/Hf 95% conf. interval
Leucosomes 0.10 1.71 0.058 0.045–0.074
Garnet-sillimanite 0.43 7.01 0.061 0.048–0.078
melanosome
accessory minerals (e.g., Bea, 1996a). Hafnium (Z = 72) is not an es-
sential structural component of any mineral. The lanthanide contrac-
tion effect causes Hf4 + and Zr4 + to have nearly identical ionic radii
(0.83 Å and 0.84 Å respectively in VIII fold coordination) so less
abundant Hf is always sequestered in Zr minerals. The Zr/Hf chondritic
ratio is 37 (McDonough and Sun, 1995) and changes little in crustal
rocks, only decreasing to Zr/Hf ≈ 25–20 in heavily fractionated leu-
cocratic segregates from granitoids (Bea et al., 2006b). Zirconium forms
some 27 minerals, 7 oxides and 20 silicates. Ubiquitous zircon (ZrSiO4)
is the most abundant, followed, albeit distantly, by baddeleyite (ZrO2),
a scarce mineral found in carbonatites and extremely silica-under-
saturated rocks. The other Zr-bearing minerals are very rare and mostly
limited to agpaitic rocks (Vlasov, 1966). Zircon contains about 1 wt% of
4
Chemical Geology xxx (xxxx) xxx–xxx
Hf and is the main Hf reservoir in common crustal rocks (Bea et al.,
2006b). Other major and accessory minerals such as titanite, amphi-
bole, clinopyroxene and garnet may also contain several ppm of Hf (see
below).
Lutetium (Z = 71) is the heaviest REE. It forms the trivalent ion
Lu3 + with an ionic radius in sixfold coordination of 0.86 Å, similar to
Y3 + (0.90 Å). Accordingly, yttrium minerals such as xenotime, yttro-
fluorite, gadolinite, kobeite, etc., and the Y-rich niobates of the fergu-
sonite-formanite, euxenite-polycrase, aeschinite, pyrochlore, and sa-
marskite groups also contain elevated concentrations of Lu and the
other HREE (Vlasov, 1966). Most of these minerals, however, are only
found in highly alkaline, often agpaitic, rocks. For this reason, the
common crustal rocks' Lu budget is shared between accessories such as
xenotime, monazite, allanite, zircon, titanite and thorite-huttonite, and
rock-forming minerals such as garnet, amphibole and clinopyroxene.
The Lu/Hf ratios of common minerals can be tentatively summar-
ized as follows:
‣ Xenotime, with 4000 to 7000 ppm Lu, has Hf ranging from 1–2 ppm
in pure xenotime to 20,000 ppm in zircon-xenotime mixed phases,
i.e., Lu/Hf from > 1000 to 0.1.
‣ Monazite contains no, or very little, Hf, and several hundred to a few
thousand ppm of Lu; accordingly, Lu/Hf is always > 1000.
‣ Apatite contains no, or very little, Hf either. Lu is highly variable,
from 0.2 to 25 ppm, albeit with the highest values resulting from
contamination with minute monazite inclusions. Lu/Hf is al-
ways > 100.
‣ Thorite-huttonite often contains several hundred to a few thousand
ppm Lu and negligible Hf, with Lu/Hf > 100.
‣ Allanite has between 15 and 50 ppm of Lu and no, or very little, Hf
with Lu/Hf generally > 100, similar to monazite and apatite.
‣ Zircon is, after xenotime and monazite, the most important Lu re-
servoir, with concentrations often around 65–90 ppm. Its elevated
Hf, however, results in Lu/Hf ≈ 0.06–0.08, the lowest values of any
mineral.
‣ Primary titanite contains somewhat elevated concentrations of Lu
(5–10 ppm) and Hf (30–50 ppm), with Lu/Hf ≈ 0.16.
‣ Amphibole often has Lu of around 0.25–0.35 ppm, and Hf 2–3 ppm,
with Lu/Hf ≈ 0.12–0.13.
‣ Clinopyroxene often has Lu ranging from 0.07 to 0.1 ppm and Hf
between 0.9 and 1.7 ppm, with Lu/Hf ≈ 0.06–0.07, i.e., about half
that of amphibole.
‣ Garnet often contains about 1.8 to 2.7 ppm Lu and 0.4 to 0.6 ppm
Hf, with Lu/Hf ≈ 3.5 to 5.7, the highest value of any silicate except
allanite.
4. Effects of Lu and Hf minerals during intracrustal processes
Not all the above-mentioned minerals play a role during magma
formation and segregation in the crust. Xenotime, for example, is pre-
sent in medium grade metapelites but disappears quickly as the tem-
perature increases, consumed in garnet-forming reactions (Bea and
Montero, 1999). The two thorium orthosilicates, huttonite and thorite,
are HREE-rich but contribute little to the Lu budget because of their
very low modal fraction. Therefore, the minerals that generally influ-
ence the Lu/Hf distribution in crustal partial melts can be reduced to (i)
monazite, apatite, zircon and garnet if the source is peraluminous, and
(ii) apatite, zircon, allanite, titanite, amphibole and clinopyroxene if the
source is metaluminous.
4.1. Lu/Hf in anatectic melts derived from felsic peraluminous protoliths
Kinetic effects aside, the dissolution of monazite, zircon and apatite
in anatectic melts depends on their intrinsic solubility and the modal
proportion of them available to the melt (Watson et al., 1989; Bea,
1996b). As melt fraction increases the relative amounts of each
F. Bea et al.
accessory “seen” by the melt are likely to approach the relative modal
proportions. If so, the melt Lu/Hf composition would be a function of
the modal proportion of the three accessories and the solubility of each
accessory. In the absence of residual garnet, the last parameter de-
termines the changes in Lu/Hf in the melt with respect to the source.
The solubility of zircon, monazite, and apatite in melts primarily
depends on temperature. Fig. 3 shows the solubility of the three mi-
nerals calculated according to the models proposed by Boehnke et al.
(2013) for zircon, Stepanov et al. (2012) for monazite, and Harrison
and Watson (1984), modified by Bea et al. (1992) for apatite. Monazite
and zircon solubility curves are nearly parallel but that of apatite is
steeper. Given that monazite and zircon exert more influence on Lumelt
and Hfmelt than apatite, it seems that the differential solubility of ac-
cessories may hardly cause significant variations in the melt Lu/Hf. To
check this, we modeled a peraluminous magmatic source using a rock
from the Peña Negra Anatectic Complex, Central Spain, as an example
(see Electronic Appendix for details). From the above-mentioned solu-
bility curves and the Winther (1995) model for estimating the compo-
sition of partial melts, we calculated Lumelt and Hfmelt assuming that no
garnet was formed and all accessories were available to the melt pro-
portional to their modal abundance. Fig. 4 shows how Lumelt and Hfmelt
increase with melt fraction (i.e., temperature) but Lu/Hfmelt stays nearly
constant with little scatter around Lu/Hfsource.
4.2. The effects of residual garnet
The effects of residual garnet and, especially, the permanence of
accessories included in residual minerals cannot be modeled as above
5
Chemical Geology xxx (xxxx) xxx–xxx
Fig. 6. Evolution of Lu and Hf and Lu/Hf as functions of the
melt fraction in metaluminous systems with no allanite or
titanite in the source (see Appendix for details). Lu/Hfmelt
decreases until zircon is exhausted from the source
(m.f. ≈ 0.3) then increases quickly until all apatite is in-
corporated into the melt (m.f. ≈ 0.45) where Lu/Hfmelt is
slightly lower than Lu/Hfsource. Afterwards Lu/Hfmelt con-
verges slowly to Lu/Hfsource as the temperature increases.
but can be determined empirically. Most granulite rocks contain zircon
included in residual garnet and orthopyroxene (Fig. 5), and this may
compensate the effects of residual garnet retention on the Lu/Hf of
segregated melts. A good example comes from the Kinzigite Formation
of Ivrea-Verbano (Mehnert, 1975; Schmid and Wood, 1976) in which
Bea and Montero (1999) studied the composition of spatially associated
granitic leucosomes and granulitic melanosomes (modal garnet >
35–40%). Despite the Lu and Hf concentrations being lower in the
leucosomes, the average Lu/Hf of these is 0.058, virtually identical to
the average Lu/Hf of the melanosomes 0.061, so indicating that no
significant variation Lu/Hf was produced despite the formation of
garnet-rich residua (Table 1). This might explain the absence of high
initial 176Hf/177Hf compositions in lower crustal granites and rhyolites
that puzzled Vervoort and Patchett (1996).
4.3. Composition of anatectic melts derived from mafic metaluminous
protoliths
In metaluminous protoliths an elevated proportion of Lu resides in
amphibole, clinopyroxene, or garnet. The rest dwells in accessories:
ubiquitous apatite and zircon, and rare allanite and titanite. This being
the case, the Lu solid/melt partitioning must be treated (i) as henrian
for the Lu that dwells in major minerals and, perhaps, apatite (ii) in
terms of accessory solubilities for zircon, allanite, and titanite. The
partitioning of Hf, on the other hand, must always be contemplated in
terms of zircon solubility. Taking into account these considerations we
modeled a metaluminous high-K calc-alkaline gabbro-dioritic source
with apatite and zircon as the only accessories. The Lumelt and Hfmelt
F. Bea et al. Chemical Geology xxx (xxxx) xxx–xxx

Fig. 7. Distribution of raw and logarithmic Lu, Hf and Lu/Hf in

0.5
1.5
A mode = 0.103
B 4784 rock analyses (see text). The red dashed line in B, D, and F
mode = 0.103 represents a gaussian distribution with the same mean and

0.4
mean = 0.348 geom. mean = 0.222 standard deviation. The three variables are lognormal so that
kernel density

kernel density
st. dev. = 0.736 their logarithms, especially ln(Lu/Hf), almost perfectly fit the
1.0

st. dev. = 0.352

0.3
skewness= 16.1 theoretical gaussian distribution. (For interpretation of the re-
skewness= 0.06 ferences to colour in this figure legend, the reader is referred to
kurtosis = 363 kurtosis = 4.57 the web version of this article.)

0.2
0.5

0.1
0.0

0.0
0.0 5.0 10.0 15.0 20.0 25.0 -6.0 -4.0 -2.0 0.0 2.0 4.0
Lu (ppm) ln(Lu)
0.00 0.01 0.02 0.03 0.04 0.05

C D 0.8
mode = 3.41
mean = 6.78 mode = 2.36
0.6
kernel density

kernel density

st. dev. = 20.34 geom. mean = 4.445

skewness= 26.4 st. dev. = 5.75
0.4

kurtosis = 863 skewness= 0.06

kurtosis = 7.16
0.2
0.0

0.0 200.0 400.0 600.0 800.0 -2.0 0.0 2.0 4.0 6.0
Hf (ppm) ln(Hf)
E
0.6

F
15

mode = 0.031
mode = 0.029
mean = 0.065
geom. mean = 0.050
kernel density

kernel density
0.4
10

st. dev. = 0.053

st. dev. = 0.056
skewness= 2.20
skewness= 0.01
kurtosis = 9.87
kurtosis = 2.81
0.2
5

0.0
0

0.0 0.1 0.2 0.3 0.4 0.5 -5.0 -4.0 -3.0 -2.0 -1.0
Lu/Hf ln(Lu/Hf)

were determined by combining the MELTS software (thermodynamics and hence little melt/restite Lu/Hf fractionation, is more the rule than
from Ghiorso and Sack, 1995; Asimow and Ghiorso, 1998; front end the exception.
from Smith and Asimow, 2005) and the above-mentioned accessory
solubility models (see details in Electronic Appendix). Fig. 6 reveals
4.4. Lu/Hf fractionation during magmatic fractionation
that Lu/Hfmelt decreases with melt fraction (m.f.) reaching a minimum
at m.f. ≈ 0.3, then increases swiftly until all apatite is exhausted from
Experimental work indicates that mafic to intermediate magmas
the source (m.f. ≈ 0.45) at which point Lu/Hfmelt is slightly lower than
become saturated in zircon at much lower temperatures than required
Lu/Hfsource. As the temperature increases further Lu/Hfmelt also in-
for the onset of crystallization of HREE-bearing rock-forming minerals
creases converging slowly to Lu/Hfsource. Little is known about the so-
such as clinopyroxene, amphibole and garnet (Hess et al., 1980;
lubility of titanite and allanite in melts, but the few existing data
Dickinson and Hess, 1982; Watson and Harrison, 1983; Boehnke et al.,
(Prince, 1943; Broska et al., 1999) indicate that both minerals' solubi-
2013). Any fractionation of such minerals occurring before zircon sa-
lity increase steeply with temperature. If so, the presence of these ac-
turation would thus decouple Lu from Hf and produce igneous rock
cessories in the source will tend to compensate the effect of zircon,
series with Lu/Hf decreasing from the most mafic to the most felsic
hindering Lu/Hf partitioning between source and melt and so leading,
members. Peraluminous granitoids, on the other hand, which in most
once more, to Lu/Hfmelt ≈ Lu/Hfsource.
cases are saturated or nearly saturated in zircon and monazite, are
If garnet remains in the source, its effects might be compensated by
likely to maintain nearly constant Lu/Hf during fractionation. Xenotime
zircon inclusions in residual minerals (see Section 4.2). Nonetheless, if
saturation, which only occurs in highly silicic melts, may increase the
there is little or no zircon, or the bulk of this mineral is available to the
Lu/Hf ratios of the most silicic members. Evidence for this is given in
melt, garnet in the residuum may cause partial melts to be notably
the next section.
depleted in Hf relative to Lu, i.e., low Lu/Hfmelt and complementary
high Lu/Hfresiduum as proposed by Vervoort and Patchett (1996).
Looking at the abundance and distribution of zircon in granulitic rocks, 5. Distribution of Lu/Hf in crustal rocks
however, it seems that the permanence of zircon inclusions in residua,
The “nugget effect” arising from residence in accessory phases

6
F. Bea et al. Chemical Geology xxx (xxxx) xxx–xxx

Table 2 variables is very sensitive to high values, outliers and mixed popula-
Averages (medians) of Lu, Hf and Lu/Hf for the whole dataset and considered groups. See tions (e.g., Yamamoto and Wakita, 2013), and is a poor estimate of the
text for explanation.
most probable value of the population. Accordingly, we preferred to
SiO2 range Lu (ppm) Hf (ppm) Lu/Hf 95% conf. interval used the median because (i) it is far less affected by outliers (ii) it is
equal to the geometric mean of the raw dataset if this is lognormally
Whole dataset (n = 4784) distributed (iii) it represents the maximum of the probability density
Total 0.222 4.445 0.050 0.049–0.051
function for the logarithmically transformed concentrations or ratios.
45–50 wt% 0.225 2.720 0.082 0.076–0.090
50–55 wt% 0.308 4.227 0.073 0.068–0.078 The averages (medians) for the whole dataset are 0.22 ppm Lu,
55–60 wt% 0.303 5.132 0.059 0.056–0.063 4.45 ppm Hf, and Lu/Hf = 0.050 (0.049–0.051 95% confidence in-
60–65 wt% 0.247 4.687 0.053 0.051–0.055 terval) (Table 2) which are quite different from the averages for the
65–70 wt% 0.195 4.698 0.041 0.040–0.043 bulk continental crust estimated by Wedepohl (1995), Rudnick and
70–74 wt% 0.184 4.621 0.040 0.038–0.041
Fountain (1995), Taylor and McLennan (1995), and Rudnick and Gao
> 74 wt% 0.207 4.509 0.046 0.042–0.050
(2003), although close to the Upper Continental Crust averages of these
M-type granitoids (n = 40)
authors. More surprisingly, they are also different from the median of
Total 0.234 1.800 0.130 0.096–0.176
45–50 wt% 0.186 1.295 0.143 0.078–0.264 granite rocks from the GEOROC database (http://georoc.mpch-mainz.
50–55 wt% 0.167 1.314 0.127 0.050–0.321 gwdg.de/georoc/) estimated by Hawkesworth et al. (2010) using 7135
55–60 wt% 0.274 2.889 0.095 0.060–0.150 samples (Lu/Hf = 0.086). The differences with the bulk continental
60–65 wt% 0.363 3.369 0.108 0.036–0.325 crust estimates may arise because (1) our data only consist of surface
65–70 wt% 0.310 3.263 0.095 0.064–0.141
70–74 wt% 0.216 2.640 0.082 0.033–0.200
samples, (2) we did not weight the representation of each rock type
> 74 wt% 0.282 2.863 0.098 0.084–0.115 according to their crustal abundance, (3) we did not include unfertile
sedimentary and metamorphic rocks such as quartzites, limestones,
I-type granitoids (n = 493)
Total 0.192 4.180 0.046 0.042–0.050 etc., (4) we reported median instead of arithmetic means. The reasons
45–50 wt% 0.266 2.146 0.124 0.097–0.157 for the discrepancy with GEOROC data are not clear. Given that this
50–55 wt% 0.350 3.603 0.097 0.072–0.131 database contains different samples from our data set, the difference
55–60 wt% 0.267 4.914 0.054 0.040–0.073
may simply reflect the large heterogeneity of Lu/Hf in crustal rocks;
60–65 wt% 0.235 4.162 0.056 0.050–0.063
65–70 wt% 0.179 4.583 0.039 0.035–0.043
however, we cannot exclude some influence of analytical artifacts due
70–74 wt% 0.116 3.973 0.029 0.025–0.034 to incomplete dissolution of zircon for ICP-MS analyses (see Section 2).
> 74 wt% 0.270 4.406 0.061 0.051–0.073 To understand whether Lu/Hf changes with magmatic fractionation,
S-type granitoids (n = 244) we studied our whole data set split in seven groups depending on the
Total 0.137 3.423 0.040 0.037–0.044 silica content. Five groups were defined at 5% intervals from SiO2 = 45
60–65 wt% 0.189 4.662 0.041 0.022–0.076 to 70 wt%: one group from 70 to 74 wt% SiO2, and the last for
65–70 wt% 0.238 4.870 0.049 0.042–0.056
SiO2 > 74 wt%. Fig. 8 shows the variation of Lu/Hf in the whole data
70–74 wt% 0.098 2.935 0.033 0.029–0.038
> 74 wt% 0.103 2.458 0.042 0.033–0.053 set by plotting the median of each group inside the 95% confidence
band. It appears that Lu/Hf varies from about 0.082 in rocks with
A-type granitoids (n = 377)
Total 1.034 13.30 0.078 0.073–0.083
45–50 wt% SiO2, to Lu/Hf ≈ 0.041 for rocks with 65–70 wt% SiO2, and
50–55 wt% 0.484 5.506 0.088 0.070–0.110 increases slightly in the most silicic samples. The figure reveals that
55–60 wt% 0.527 13.392 0.039 0.022–0.071 using the mafic and felsic source averages may cause large errors; it also
60–65 wt% 0.795 13.252 0.060 0.053–0.068 suggests that using a single fixed Lu/Hfsource value similar to the Upper
65–70 wt% 0.866 13.608 0.064 0.050–0.081
Continental Crust average (as used by Bodet and Schärer, 2000; and
70–74 wt% 1.276 17.400 0.073 0.067–0.080
> 74 wt% 1.053 11.453 0.092 0.082–0.103 Iizuka et al., 2005) is likely to yield the most accurate TDM Hf estimates.
Then, to characterize the Lu/Hf distribution in different rock series
TTG (n = 304)
Total 0.100 3.601 0.028 0.026–0.030 we divided each silica group in six subsets depending on the rock ty-
60–65 wt% 0.141 3.601 0.039 0.033–0.047 pology: M-type, I-type, S-type, A-type, Archean TTGs and adakites, note
65–70 wt% 0.112 3.894 0.029 0.026–0.032 that this division neither attempts to be systematic, exhaustive or
70–74 wt% 0.086 3.414 0.025 0.022–0.029 geochemically correct, nor pretends to legitimize these categories. We
> 74 wt% 0.058 3.106 0.019 0.014–0.025
simply used it as a means to investigate whether magmatic rocks
Adakitic rocks (n = 602) coming from different sources show meaningful differences in the Lu/
Total 0.130 2.963 0.044 0.041–0.047
Hf distribution. It must also be considered that the ascription of samples
45–50 wt% 0.273 1.459 0.187 0.138–0.253
50–55 wt% 0.281 2.283 0.123 0.090–0.168 to a given category is the one made by the authors, and only when this
55–60 wt% 0.189 3.566 0.053 0.043–0.066 is clearly stated; doubtful or unclear cases were not considered.
60–65 wt% 0.146 3.184 0.046 0.043–0.049 Accordingly only 2065 out of 4784 samples were ascribed to groups.
65–70 wt% 0.108 3.234 0.033 0.031–0.036 Fig. 9 and Table 2 show the differences between the rock types con-
70–74 wt% 0.089 3.175 0.028 0.025–0.032
> 74 wt% 0.146 2.907 0.050 0.033–0.075
sidered. M-type granitoids display the highest Lu/Hf, with averages
close to the generally recommended values; however they also show a
tremendous variation, likely caused by the scarce number of samples
causes the distributions of Lu and Hf to be markedly asymmetric to- included in this group (Table 2). The I-type granitoids have Lu/Hf
wards their highest values, with elevated skewness of 16.1 and 26.4, around 0.124 for the less silicic rocks, decreasing to about 0.056 for
respectively (Fig. 7A and C). Logarithmic concentrations, on the other samples with 55–60 wt% SiO2 and 0.03 for samples with 70–74 wt%
hand, yield nearly normal low-skewness distributions with some excess SiO2 then increasing to 0.06 in the most silicic samples. The S-type
kurtosis (Fig. 7B and D). The distribution of individual Lu/Hf ratios is granitoids and anatexites have nearly constant values around Lu/
also skewed (Fig. 7E), albeit not so much as the individual elements, Hf ≈ 0.04. The A-type granitoids have values from Lu/Hf ≈ 0.07 for
and the logarithm of Lu/Hf fits a neat normal distribution (Fig. 7F). The rocks in the 50–55 wt% SiO2 range to Lu/Hf ≈ 0.09 at the silicic end.
three distributions are, therefore, lognormal. This poses various theo- The Archean TTGs show the lowest Lu/Hf ratios of all types of grani-
retical and practical problems when calculating the best average of the toids, in most cases around Lu/Hf ≈ 0.03. Lastly, the rocks classified as
populations because the arithmetic mean of the lognormally distributed adakites have Lu/Hf = 0.187 for 45–50 wt% SiO2; the ratio decreases
to Lu/Hf = 0.053 for 50–55 wt% SiO2 and Lu/Hf = 0.028 for 70–74 wt

7
F. Bea et al. Chemical Geology xxx (xxxx) xxx–xxx

0.20
% SiO2, and increases slightly in samples with SiO2 > 75 wt%.

6. Influence on zircon TDM Hf calculations

0.15

To check the correctness of the selected Lu/Hf value, in this section

Lu/Hf
0.10

we compared the whole rock Nd and zircon Hf model ages. It should be

noted that the Nd model ages do not need to be accurate estimates of
mantle extraction ages; they can also represent mixed ages. If the par-
0.05

ticular Lu/Hf used in the zircon Hf model age calculation truly re-
presents the source composition, the two model ages will approximate
one another thus indicating that they reflect similar source components
0.00

and that these mixed in similar proportions.

50 55 60 65 70 75
SiO2 (wt.%) 6.1. Fixed Lu/Hf crustal averages versus averages calculated depending on
Fig. 8. Average and 95% confidence band of Lu/Hf vs SiO2 for the whole data set. Rocks SiO2
with SiO2 < 50 wt% cluster around Lu/Hf = 0.082; rocks with SiO2 between 60 and
65 wt% cluster around 0.053. The two values are much lower than the normally assumed The medians calculated here are markedly different from the com-
mafic (M.S.) and felsic source averages (F.S.), but close to the upper continental crust monly used “mafic and felsic magmatic sources” Lu/Hf averages. Only
average (U.C.C. ≈ 0.05; e.g., Rudnick and Gao, 2003).
the less silicic adakites (< 50 wt% SiO2) have Lu/Hf comparable to, or
even higher than, the “average mafic source” (Lu/Hf = 0.165); all
other rocks with SiO2 > 55 wt% have markedly lower Lu/Hf (Table 2).
Similarly, most rocks with SiO2 between 60 and 74 wt%, which are the

PONER NUMERO DE JUESTRAS 0.20

0.20

0.15
0.15
Lu/Hf

0.10
0.10

0.05
0.05

M-type I-type
0.00
0.00

50 55 60 65 70 75 50 55 60 65 70 75
0.20

0.20

SiO2 SiO2
0.15

0.15
Lu/Hf
0.10

0.10
0.05

0.05

S-type A-type
0.00

0.00

50 55 60 65 70 75 50 55 60 65 70 75
0.20

0.20

SiO2 SiO2
0.15

0.15
Lu/Hf
0.10

0.10
0.05

0.05

archean TTG adakites

0.00

0.00

50 55 60 65 70 75 50 55 60 65 70 75

SiO2 (wt.%) SiO2 (wt.%)

Fig. 9. Average and 95% confidence band of Lu/Hf vs SiO2 for different rock series. The large dispersion of M-type granitoids is caused by the scarce number of data included in this
category (Table 1). See text for explanation.

8
F. Bea et al. Chemical Geology xxx (xxxx) xxx–xxx

3500

1600
F.S. M.S. A) Lu/Hf analytically determined

1400
it e
3000

zircon TDM Hf (Ma)

n
ra
Hf model age (TDM Hf)

z og
on

1200
am
2500

fr om ro
ns bb
r co a
a zi gg

1000
05 M - M
6 h
2000

TDM Nd hig ge
s
ma la
s fro a
a zir
con qu

800
e
313 M
1500

TDM Nd 800 1000 1200 1400 1600

TCRNd (Ma)

1600
1000

0.03 0.06 0.09 0.12 0.15 0.18 B) Lu/Hf estimated from SiO2

1400
Lu/Hf (element ratio)

zircon TDM Hf (Ma)

Fig. 10. Changes in Hf model age depending on the Lu/Hf assumed for the source for two

1200
real zircon populations: one from a Variscan high-Mg gabbro of central Spain
(SiO2 = 52.3 wt%); and the other from a Pan-African granite of western Egypt
(SiO2 = 70.6 wt%). F.C. and M.C. represent the felsic crust and mafic crust averages,

1000
respectively. To obtain zircon Hf model ages similar to the whole-rock Nd model ages it is
necessary to assume a Lu/Hf = 0.055, for the granite and Lu/Hf = 0.083 for the gabbro. es
l ag
These values are much lower than F.C. (Lu/Hf = 0.09) and M.C. (Lu/Hf = 0.165) but
ua
eq

800
almost exactly match the data set average for the 60–65 and 45–50 wt% SiO2 intervals,
respectively. 800 1000 1200 1400 1600
TCRNd (Ma)
most probable source for crustal granites, have Lu/Hf of 0.05 or less,
notably lower than the “average felsic magmatic source” used by most 2500
authors. The only exception is M-type granitoids, which have Lu/ C) Lu/Hf = 0.165 (”mafic source”)
Hf ≈ 0.08–0.10 although with notable dispersion (Fig. 9).
2000
zircon TDM Hf (Ma)

To check whether the Lu/Hf ratios calculated here as a function of

rock's silica concentration result in more accurate Hf model ages than
1500

the “mafic and felsic magmatic source” averages, we compared the

zircon TDM Hf to the whole rock TDM Nd (see parameters in Section 2) in
ages
1000

equal
two cases of very different rocks: one Variscan high-Mg gabbro from the
Arenal intrusive complex, central Spain (Bea et al., 1999), and a Pan-
African granite from Bir Safsaf, Egyptian Western Desert (Bea et al.,
500

2011).
Fig. 10 shows the variations of the mean TDM Hf of the two zircon
populations as a function of the Lu/Hf assumed for the calculation. In
the case of the high-Mg gabbro the mean zircon Hf model age calcu-
lated assuming Lu/Hf = 0.165 is TDM Hf = 2.08 Ga, about 700 Myr
higher than the Nd model age. However, using Lu/Hf = 0.083
(equivalent to 176Lu/177Hf = 0.0116) which is nearly equal to the
whole dataset median for the 45–50 wt% SiO2 range (Table 2), the
zircon Hf model age comes close to the Nd model age. Similarly, the
mean zircon Hf model age of the Bir Safsaf granite calculated assuming
the Lu/Hf value for a felsic crust (0.09) is TDM Hf = 2.15 Ga, i.e., about
300 Myr higher than the Nd model age. Again, to obtain equal Nd and
Hf model ages requires a lower value of Lu/Hf = 0.055 (equivalent to
176
Lu/177Hf = 0.0076) which matches the calculated rock average for
the 60–65 wt% silica range (Lu/Hf = 0.053, Table 2).
6.2. Fixed averages versus whole-rock Lu/Hf
Another example that may indicate the best strategy for selecting
the most appropriate Lu/Hf values is illustrated in Fig. 11 which
compares different calculations of TDM Hf for a suite of Iberian mafic to
intermediate rocks to whole-rock Nd model ages. Studied samples are
seven gabbros (SiO2 from 47.6 to 52.2 wt%), one quartzdiorite
(SiO2 = 59.3 wt%), and one tonalite (SiO2 = 65.6 wt%); all of them
have uniform zircon populations the ages of which range between 308
and 315 Ma (Montero et al., 2004; Bea et al., 2006a). Fig. 11A plots the
zircon TDM Hf calculated from the analytically-determined whole-rock
Lu/Hf against the whole-rock Nd model ages; all data scatter around the
800 1000 1200 1400 1600
TCRNd (Ma)
Fig. 11. Different calculations of zircon TDM Hf for a suite of Iberian mafic to intermediate
rocks compared to whole-rock Nd model ages. Analytically determined whole-rock Lu/Hf
ratios produce the best coincidence. Lu/Hf estimated from the rock SiO2 concentration
produces worse, yet acceptable, results, but those using the “average mafic crust” seem
unacceptable.
equality line and the maximum discrepancies are ~100 Myr. Fig. 11B
plots the TDM Hf calculated from the Lu/Hf ratio estimated according to
the rock SiO2 concentration as shown in Fig. 8. The two ages also
scatter around the equality line but the discrepancies increase up to
200 Myr. Fig. 11C plots the TDM Hf calculated from the “mafic crust
value” of Lu/Hf = 0.165. In this case all zircon Hf model ages are
greatly overestimated with respect to the Nd model ages, between 400
and 700 Myr. Despite common belief, therefore, the best agreement
between zircon Hf and whole-rock Nd model ages arises when the
analytically determined whole-rock Lu/Hf ratio is used in the calcula-
tions, at least in this case. Using the Lu/Hf estimated from the rock SiO2
concentration produces worse, yet acceptable, results, but those cal-
culated using the “average mafic crust” seem unacceptable.
6.3. Effects of fixed averages on rocks generated from heterogenous sources
The large Lu/Hf variation between the different granite types
strongly suggests that crustal segments involved in orogenesis and
anatexis are likely to be heterogeneous with respect to Lu/Hf. To

9
F. Bea et al.
kernel density (arbitrary units)
melting at
300 Ma
1500 2000
Hf model age (Ma)
kernel density (arbitrary units)
melting at
1000 Ma
1500 2000
Hf model age (Ma)
kernel density (arbitrary units)
melting at
2000 Ma
1500 2000
Hf model age (Ma)
Fig. 12. Simulation of the distribution of Hf model ages in zircons crystallized from
magmas generated at different times from a 2500 Ma source that was bimodal with re-
spect to Lu/Hf. TDM Hf was calculated assuming a fixed Lu/Hfsource = 0.05. Note how the
Lu/Hf bimodality of the source translates in a bimodal distribution of TDM Hf that might
be erroneously interpreted as reflecting two episodes of crustal growth. See text for ex-
planation.
understand the effects of assuming a fixed Lu/Hfsource value on the Hf
model ages of zircons crystallized in magmas derived from hetero-
geneous sources, we undertook the following simulation (see details in
the Electronic Appendix). Let us consider a crustal segment consisting
of different igneous rocks generated at 2.5 Ga from an isotopically
homogeneous source with 176Hf/177Hf = 0.281439 and εHf = 8.9, that
is, close to the depleted mantle at that time. The magmas these 2.5 Ga
rocks crystallized from were thus initially isotopically homogeneous but
had different Lu/Hf as a result of fractionation processes during partial
melting and emplacement. For simplicity we can assume that the 2.5 Ga
rocks were bimodal, diorites and granites, with Lu/Hf ≈ 0.092 and Lu/
Hf ≈ 0.05, respectively. These rocks remained stable in the crust until
the Early Carboniferous, when they were eroded producing greywacke-
like sediments that were metamorphosed and melted at 300 Ma to yield
anatectic granites with newly crystallized zircons.
Given that the chemical composition of anatectic granites derived
from metagreywackes would be broadly similar, i.e., peraluminous
monzogranites or leucogranites (e.g., Montel and Vielzeuf, 1997), there
is no reasons to presume they ultimately came from two sources with
Chemical Geology xxx (xxxx) xxx–xxx
different Lu/Hf. Accordingly, we used a single fixed Lu/Hfsource value
A for all zircons separated from the granites, choosing Lu/Hf = 0.05 be-
cause the source is likely to be felsic. Fig. 12 A reveals that doing so
results in a beautiful bimodal TDM Hf distribution with maxima at
2.10 Ga and 2.49 Ga, respectively. If we displace the anatectic event to
1.0 Ga (Fig. 12B) or 2.0 Ga (Fig. 12C), the resulting TDM Hf distributions
are also bimodal but the two modes become closer as the age of anatexis
approaches that of the source, merging when the two ages are equal.
Remarkably, the bimodality appears independently of the particular
Lu/Hfsource assumed; increasing this parameter simply modifies the
position of the two modes towards higher ages and slightly increases
2500 3000
distance between them.
Most geologists would be tempted to interpret Fig. 12 as evidence
for two different pulses of crustal growth in this area, hardly imagining
that it reflects a bimodal source formed during a single event. This il-
B lustrates how using a fixed Lu/Hfsource for calculating zircon Hf model
ages risks translating the Lu/Hf heterogeneity of the source into a TDM
Hf distribution that may be erroneously interpreted as different epi-
sodes of crustal growth, in particular when dealing with zircons much
younger than their sources.
7. Conclusions
Crustal rocks have highly variable Lu/Hf depending on the rock
2500 3000 type and SiO2 concentration. The highest Lu/Hf ratios are found in
rocks with elevated proportion of juvenile mantle components, M-type
granitoids and adakites, and the lowest in rocks with a protracted
crustal history such as S-type granitoids and Archean TTGs. Except for
A-type granitoids, the highest Lu/Hf of each rock type occur in the
C lowest SiO2 range. Lu/Hf decreases, and the discrepancies between the
rock series attenuate, as SiO2 increases. The most silicic members of
each rock series (SiO2 > 73 wt%) often show a small increase in Lu/Hf
which usually is related to the presence of modal xenotime.
The Lu/Hf distribution in 4784 igneous rocks of different ages from
all over the world is lognormal rather than normal. Accordingly, the
median instead of the arithmetic mean should be used for estimating
rock averages if these are to reflect the maximum of the probability
density function. The Lu/Hf averages for potential mafic and felsic
2500 3000
magmatic sources calculated in this way are Lu/Hfmafic ≈ 0.08 and Lu/
Hffelsic ≈ 0.05, much lower than customarily assumed in the calcula-
tions of the Hf model ages: Lu/Hfmafic ≈ 0.165 and Lu/Hffelsic ≈ 0.09.
This may lead to huge overestimations of the zircon TDM Hf, especially
when there is a large difference between zircon and source ages.
Partial melting of peraluminous sources causes little decoupling of
Lu from Hf because the solubility curves of monazite and zircon with
increasing temperature are nearly parallel. Partial melting of metalu-
minous sources may decouple Lu from Hf at low melt fractions, espe-
cially if allanite and/or titanite are not present because the elevated
solubility of zircon in hot low-silica melts may produce lower-than-
source Lu/Hfmelt if the melt fraction is small, but Lu/Hfmelt quickly ap-
proaches Lu/Hfsource as the melt fraction increases. The presence of al-
lanite or titanite in the source results in Lu/Hfmelt similar to Lu/Hfsource
irrespective of the melt fraction. The effect of residual garnet on Lu/Hf
melt/residuum partitioning is counterbalanced by the retention of
zircon included in garnet and other residual minerals.
Fractional crystallization of peraluminous magmas hardly frac-
tionates Lu/Hf because these magmas were either formed already sa-
turated in zircon and monazite or became saturated in the two minerals
at similar temperatures. Fractional crystallization of metaluminous
magmas may decouple Lu from Hf if amphibole or clinopyroxene, in
which HREEs are compatible, begin crystallizing before zircon satura-
tion and the melt-solid segregation is efficient. If so, the less silicic rocks
of the series will have higher Lu/Hf than the most silicic rocks.
Calculating Hf model ages using a fixed Lu/Hfsource value risks
translating Lu/Hfsource heterogenities into heterogeneities in the re-
sulting TDM Hf, and these can be erroneously interpreted as evidence for
10
F. Bea et al.
different pulses of crustal growth.
For these reasons, and based on the above considerations, we sug-
gest that the best strategy to calculate TDM Hf is to use the analytically
determined whole-rock Lu/Hf ratio as a proxy for the source Lu/Hf. The
rock type and SiO2 concentration dependent averages shown in Table 2
also are a good approximation if the source was homogeneous with
respect to Lu/Hf. For detrital or inherited zircons, for which no whole-
rock information is available, it seems advisable to determine first
whether they crystallized in mafic or felsic rocks by interpreting cath-
odoluminescence images, Th/U ratios, etc.…, and then estimate the Lu/
Hfsource from the appropriate average in Table 2.
Acknowledgments
The authors are indebted to T. Iizuka and two anonymous referees
for their helpful and constructive revisions. This paper has been fi-
nanced by the grants CGL2013-40785-P, Ministerio de Economía y
Competitividad de España, and P12.RNM.2163, Junta de Andalucía.
This is the IBERSIMS publication n° 48.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://
doi.org/10.1016/j.chemgeo.2017.11.034.
References
Amelin, Y., Lee, D.C., Halliday, A.N., Pidgeon, R.T., 1999. Nature of the Earth's earliest
crust from hafnium isotopes in single detrital zircons. Nature 399 (6733), 252–255.
Asimow, P.D., Ghiorso, M.S., 1998. Algorithmic modifications extending MELTS to cal-
culate subsolidus phase relations. Am. Mineral. 83, 1127–1131.
Bea, F., 1996a. Residence of REE, Y, Th and U in granites and crustal protoliths; im-
plications for the chemistry of crustal melts. J. Petrol. 37, 521–552.
Bea, F., 1996b. Controls on the trace element composition of crustal melts. Trans. R. Soc.
Edinb. Earth Sci. 87, 33–42.
Bea, F., Montero, P., 1999. Behavior of accessory phases and redistribution of Zr, REE, Y,
Th, and U during metamorphism and partial melting of metapelites in the lower crust:
an example from the Kinzigite Formation of Ivrea-Verbano, NW Italy. Geochim.
Cosmochim. Acta 63, 1133–1153.
Bea, F., Fersthater, G., Corretgé, L.G., 1992. The geochemistry of phosphorus in granite
rocks and the effect of aluminium. Lithos 29, 43–56.
Bea, F., Montero, P., Molina, J.F., 1999. Mafic precursors, peraluminous granitoids, and
late lamprophyres in the Avila batholith: a model for the generation of Variscan
batholiths in Iberia. J. Geol. 107, 399–419.
Bea, F., Montero, P., Zinger, T.F., 2003. The nature and origin of the granite source layer
of Central Iberia: evidence from trace element, Sr and Nd isotopes, and zircon age
patterns. J. Geol. 111, 579–595.
Bea, F., Montero, P., González-Lodeiro, F., Talavera, C., Molina, J.F., Scarrow, J.H.,
Whitehouse, M.J., Zinger, T.F., 2006a. Zircon thermometry and U-Pb ion-microprobe
dating of the gabbros and associated migmatites of the Variscan Toledo Anatectic
Complex, Central Iberia. J. Geol. Soc. Lond. 163, 847–855.
Bea, F., Montero, P., Ortega, M., 2006b. A LA-ICPMS evaluation of Zr reservoirs in
common crustal rocks: implications for Zr and Hf geochemistry, and zircon-forming
processes. Can. Mineral. 44, 693–714.
Bea, F., Montero, P., Abu Anbar, M., Molina, J.F., Scarrow, J.H., 2011. The Bir Safsaf
Precambrian inlier of South West Egypt revisited. A model for ~ 1.5 Ga TDM late Pan-
African granite generation by crustal reworking. Lithos 125, 897–914.
Blichert-Toft, J., Albarede, F., 1997. The Lu-Hf isotope geochemistry of chondrites and
the evolution of the mantle-crust system. Earth Planet. Sci. Lett. 148, 243–258.
Blichert-Toft, J., Albarède, F., 2008. Hafnium isotopes in Jack Hills zircons and the for-
mation of the Hadean crust. Earth Planet. Sci. Lett. 265, 686–702.
Bodet, F., Schärer, U., 2000. Evolution of the SE-Asian continent from U-Pb and Hf iso-
topes in single grains of zircon and baddeleyite from large rivers. Geochim.
Cosmochim. Acta 64, 2067–2091.
Boehnke, P., Watson, E.B., Trail, D., Harrison, T.M., Schmitt, A.K., 2013. Zircon saturation
re-revisited. Chem. Geol. 351, 324–334.
Bouvier, A., Vervoort, J.D., Patchett, P.J., 2008. The Lu–Hf and Sm–Nd isotopic compo-
sition of CHUR: constraints from unequilibrated chondrites and implications for the
bulk composition of terrestrial planets. Earth Planet. Sci. Lett. 273, 48–57.
Broska, I., Chekmir, A.S., Hatar, J., 1999. Allanite solubility and the role of accessory
mineral paragenesis in the Carpatian granite petrology. Geol. Carpath. 50, SI90–91.
Dhuime, B., Hawkesworth, C.J., Storey, C.D., Cawood, P.A., 2011. From sediments to
their source rocks: Hf and Nd isotopes in recent river sediments. Geology 39,
407–410.
Dickinson, J.E., Hess, J.C., 1982. Zircon saturation in lunar basalts and granites. Earth
Planet. Sci. Lett. 57, 336–344.
Ghiorso, M.S., Sack, R.O., 1995. Chemical mass transfer in magmatic processes IV. A
11
Chemical Geology xxx (xxxx) xxx–xxx
revised and internally consistent thermodynamic model for the interpolation and
extrapolation of liquid solid equilibria in magmatic systems at elevated temperatures
and pressures. Contrib. Mineral. Petrol. 119, 197–212.
Goldstein, S.L., O'Nions, R.K., Hamilton, P.J., 1984. A Sm-Nd isotopic study of atmo-
spheric dust and particulates from major river systems. Earth Planet. Sci. Lett. 70,
221–236.
Griffin, W.L., Wang, X., Jackson, S.E., Pearson, N.J., O'Reilly, S.Y., Xu, X.S., Zhou, X.M.,
2002. Zircon chemistry and magma mixing, SE China: in-situ analysis of Hf isotopes,
Tonglu and Pingtan igneous complexes. Lithos 61, 237–269.
Hall, G.E.M., Plant, J.A., 1992. Analytical errors in the determination of high field
strength elements and their implications in tectonic interpretation studies. Chem.
Geol. 95, 141–156.
Harrison, T.M., Watson, E.B., 1984. The behavior of apatite during crustal anatexis:
equilibrium and kinetic considerations. Geochim. Cosmochim. Acta 48, 1467–1478.
Hawkesworth, C.J., Kemp, A.I.S., 2006. Using hafnium and oxygen isotopes in zircons to
unravel the record of crustal evolution. Chem. Geol. 226, 144–162.
Hawkesworth, C.J., Dhuime, B., Pietranik, A.B., Cawood, P.A., Kemp, A.I.S., Storey, C.D.,
2010. The generation and evolution of the continental crust. J. Geol. Soc. 167,
229–248.
Hess, J.C., Dickinson, J.E., Rutherford, M.J., 1980. Solubility of zircon, whitlockite and
apatite in lunar basalts and granites. Lunar Planet. Sci. 11, 438–440.
Iizuka, T., Hirata, T., Komiya, T., Rino, S., Katayama, I., Motoki, A., Maruyama, S., 2005.
U-Pb and Lu-Hf isotope systematics of zircons from the Mississippi River sand: im-
plications for reworking and growth of continental crust. Geology 33, 485.
Iizuka, T., Yamaguchi, T., Itano, K., Hibiya, Y., Suzuki, K., 2017. What Hf isotopes in
zircon tell us about crust–mantle evolution. Lithos 274–275, 304–327.
Jacobsen, S.B., Wasserburg, G.J., 1980. Sm-Nd isotopic evolution of chondrites. Earth
Planet. Sci. Lett. 50, 139–155.
Jacobsen, S.B., Wasserburg, G.J., 1984. Sm-Nd isotopic evolution of chondrites and
achondrites, II. Earth Planet. Sci. Lett. 67, 137–150.
Kemp, A.I.S., Hawkesworth, C.J., Foster, G.L., Paterson, B.A., Woodhead, J.D., Hergt,
J.M., Gray, C.M., Whitehouse, M.J., 2007. Magmatic and crustal differentiation his-
tory of granitic rocks from Hf-O isotopes in zircon. Science 315, 980–983.
Kretz, R., 1983. Symbols for rock-forming minerals. Am. Mineral. 68, 277–279.
McDonough, W.F., Sun, S.S., 1995. The composition of the earth. Chem. Geol. 120,
223–253.
Mehnert, K.R., 1975. The Ivrea Zone, a model of the deep crust. Neues Jb. Mineral. Abh.
125, 156–199.
Montel, J.M., Vielzeuf, D., 1997. Partial melting of metagreywackes, part II. Compositions
of minerals and melts. Contrib. Mineral. Petrol. 128, 176–196.
Montero, P., Bea, F., Zinger, T.F., 2004. Edad 207Pb/206Pb en cristal único de circón de las
rocas máficas y ultramáficas del sector de Gredos, Batolito de Avila (Iberia Central).
Revista de la Sociedad Geológica Española 17, 157–165.
Moyen, J.-F., Martin, H., 2012. Forty years of TTG research. Lithos 148, 312–336.
Park, C.-S., Shin, H.S., Oh, H., Moon, J.H., Cho, H., Cheong, C.-s., 2013. Determination of
trace elements in geological reference materials G-3, GSP-2 and SGD-1a by low-di-
lution glass bead digestion and ICP-MS. Geostand. Geoanal. Res. 37, 361–368.
Patchett, P.J., Vervoort, J.D., Söderlund, U., Salters, V.J.M., 2004. Lu–Hf and Sm–Nd
isotopic systematics in chondrites and their constraints on the Lu–Hf properties of the
Earth. Earth Planet. Sci. Lett. 222, 29–41.
Payne, J.L., McInerney, D.J., Barovich, K.M., Kirkland, C.L., Pearson, N.J., Hand, M.,
2016. Strengths and limitations of zircon Lu-Hf and O isotopes in modelling crustal
growth. Lithos 248-251, 175–192.
Prince, A.T., 1943. The system albite-anorthite-sphene. J. Geol. 51, 1–16.
Rudnick, R.L., Fountain, D.M., 1995. Nature and composition of the continental-crust - a
lower crustal perspective. Rev. Geophys. 33, 267–309.
Rudnick, R.L., Gao, S., 2003. Composition of the continental crust. In: Rudnick, R.L. (Ed.),
The Crust. Treatise in Geochemistry 3. pp. 1–64.
Scherer, E., Münker, C., Mezger, K., 2001. Calibration of the lutetium-hafnium clock.
Science 293, 683–687.
Schmid, R., Wood, B.J., 1976. Phase relationships in Granulitic Metapelites from the
Ivrea-Verbano Zone (Northern Italy). Contrib. Mineral. Petrol. 54, 255–279.
Smith, P.M., Asimow, P.D., 2005. Adiabat_1ph: a new public front-end to the MELTS,
pMELTS and pHMELTS models. Geochem. Geophys. Geosyst. 6, Q02004. http://dx.
doi.org/10.1029/2004GC000816. (art. no).
Söderlund, U., Patchett, P.J., Vervoort, J.D., Isachsen, C.E., 2004. The 176Lu decay con-
stant determined by Lu–Hf and U–Pb isotope systematics of Precambrian mafic in-
trusions. Earth Planet. Sci. Lett. 219, 311–324.
Stepanov, A.S., Hermann, J., Rubatto, D., Rapp, R.P., 2012. Experimental study of
monazite/melt partitioning with implications for the REE, Th and U geochemistry of
crustal rocks. Chem. Geol. 300–301, 200–220.
Taylor, S.R., McLennan, S.M., 1995. The geochemical evolution of the continental crust.
Rev. Geophys. 33, 241–265.
Thirlwall, M.F., Walder, A.J., 1995. In situ hafnium isotope ratio analysis of zircon by
inductively coupled plasma multiple collector mass spectrometry. Chem. Geol. 122,
241–247.
Totland, M., Jarvis, I., Jarvis, K.E., 1992. An assessment of dissolution techniques for the
analysis of geological samples by plasma spectrometry. Chem. Geol. 95, 35–62.
Vervoort, J.D., Kemp, A.I.S., 2016. Clarifying the zircon Hf isotope record of crust–mantle
evolution. Chem. Geol. 425, 65–75.
Vervoort, J.D., Patchett, P.J., 1996. Behavior of hafnium and neodymium isotopes in the
crust: constraints from Precambrian crustally derived granites. Geochim. Cosmochim.
Acta 60, 3717–3733.
Vlasov, K.A., 1966. Geochemistry and Mineralogy of Rare Elements and Genetic Types of
Their Deposits. Volume II: Mineralogy of Rare Elements. Israel Program for Scientific
Translation, Jerusalem (939 pp).
F. Bea et al.
Watson, E.B., Harrison, T.M., 1983. Zircon saturation revisited: temperature and com-
position effects in a variety of crustal magma types. Earth Planet. Sci. Lett. 64,
295–304.
Watson, E.B., Vicenzi, E.P., Rapp, R.P., 1989. Inclusion/host relations involving accessory
minerals in high-grade metamorphic and anatectic rocks. Contrib. Mineral. Petrol.
101, 220–231.
Wedepohl, K.H., 1995. The composition of the continental crust. Geochim. Cosmochim.
Acta 59, 1217–1232.
12
Chemical Geology xxx (xxxx) xxx–xxx
Winther, K.T., 1995. A model for estimating the composition of partial melts. Mineral.
Petrol. 53, 189–195.
Woodhead, J., Hergt, J., Shelley, M., Eggins, S., Kemp, R., 2004. Zircon Hf-isotope ana-
lysis with an excimer laser, depth profiling, ablation of complex geometries, and
concomitant age estimation. Chem. Geol. 209, 121–135.
Yamamoto, K., Wakita, J.-I., 2013. “Average-Value Paradox” of the lognormal distribu-
tion. J. Phys. Soc. Jpn. 82, 113001 (art. n° 3 pages).